Download Ling 107 Syntax - The Study of Sentence Structure All human

Ling 107
Syntax - The Study of Sentence Structure
All human languages use sentences as minimal units of propositional expression, but the forms
sentences can take in any language are infinitely varied. The study of sentence structure exposes
the way in which human creativity is constrained by structure.
Compare the possibilities for creation in the lexicon and with sentences:
What is the longest word you know? / The longest sentence?
Can you make up a new word? / A new sentence?
Both questions reveal a level of creativity available at the sentence level, but not at the lexical
level of a language. Syntax is the linguist’s attempt to understand this creativity and its limits.
We know there are limits on the sentences we produce just as there are limits on the forms of
words in any language. Any speaker of English would agree that “* Take me out game the ball”
is not an acceptable sentence. (Linguists use asterisks to mark unacceptable sentences.) How can
we produce an infinite number of original sentences on one hand, and recognize that an infinite
number of sentences are unacceptable on the other?
Chomsky (1965) observed that the creativity we observe in sentence construction raises an
interesting problem for language acquisition. Do children learn how to create sentences?
One hypothesis would be that children learn a language by simply imitating the sentences that
other speakers produce. Imitation does not explain children’s ability to produce new sentences, or
accept sentences that they have never heard other speakers produce.
A second hypothesis would be that children produce new sentences by combining words together
in novel ways. Random generation would predict that children would also produce lots of
unacceptable sentences before they learn the correct structure for sentences.
A third hypothesis is that children produce new sentences based on analogy with the sentences
they hear other speakers produce. Analogy predicts children would also produce false analogies,
e.g., *Are the girls who ____ good are getting cake?
from Are the girls getting cake?
The solution to the acquisition problem lies in recognizing sentence structure and role it plays in
organizing words.
Sentence Structure
An account of sentence structure has to explain two basic features of sentences:
1. The linear order of the words in the sentence.
A cat is on the mat =/ The mat is on the cat.
2. The grouping of words into constituents.
Constituent structure is another fundamental aspect of our grammatical knowledge. Not only can
we recognize the difference between acceptable and unacceptable sentences, we can parse any
acceptable sentence into smaller phrases or constituents, e.g.,
[Any speaker of English]
[can split this sentence into two main constituents]
We can take this process further:
[Any] [speaker of English]
[can]
[split this sentence into two main constituents]
And even further if we want to.
Linguists use constituency tests as evidence to back up their intuitions about sentence
components. Ideally the different constituency tests produce the same results, but they can also
produce different results. It takes some practice with the tests to learn how to use them properly
and interpret their results.
1. Semantic Intuition Test—decide if the words can be grouped together semantically.
Your intuitions about the semantic relations between the words in a sentence provide a basic
insight into a sentence’s constituent structure.
The phrase ‘into two main constituents’ specifies a result and is therefore a candidate
constituent.
2. Substitution Test—see if a pro-form can be substituted for the proposed constituent.
Pronouns are an example of a pro-form, a form that can be substituted for a constituent of a
specific type:
pronouns
substitute for
determiner phrases
do
substitutes for
verb phrases
one
substitutes for
noun phrases
there
substitutes for
prepositional phrases
auxiliaries
substitute for
auxiliary phrases
3. Stand Alone Test—see if the proposed constituent can stand by itself.
It helps to imagine using the constituent by itself in response to a question:
Who can split this sentence into two main constituents?
Any speaker of English.
How can any speaker of English split this sentence?
Into two main constituents.
4. Movement Test—see if the proposed constituent can be moved.
The movement test has more limitations than the other tests and is correspondingly more difficult
to apply. One trick is to use a pseudocleft construction:
[Two main constituents] is what [any speaker of English can split this sentence into].
as compared with
? Two main constituents any speaker of English can split this sentence into.
You should learn how to use the constituency tests to show that a set of words is NOT a
constituent as well as to show that it is, e.g.,
1. Semantic Intuition
‘this sentence into two main constituents’ does not make sense.
2. Substitution
Any speaker of English can split this =/
Any speaker of English can split this sentence into two main constituents
3. Stand Alone
What can any speaker of English split?
*This sentence into two main constituents
4. Movement
*This sentence into two main constituents is how any speaker of English can split.
Constituent Properties
Once we agree on the existence of sentence constituents, we can explore how they are used to
construct sentences.
We can identify different types of constituents, or phrases, e.g.,
Noun phrases: any speaker of English, this sentence, two main constituents
Verb phrases: split this sentence into two main constituents
Prepositional phrases: into two main constituents
The different types of phrases have some basic similarities.
One word, the head, is central to every type of phrase.
The head can take a complement phrase that provides an argument for the head, e.g.,
NP
ru
speaker of English
VP
ru
split this sentence
PP
ru
into two main constituents
Each phrase can take a specifier that makes the referent more explicit, e.g.,
NP
ru
any
speaker of English
VP
ru
can
split
PP
ru
only
into two
Abstracting away from the specifics, we can construct a template for ALL phrases:
Phrase
u
Intermediate
ru
Specifier
Head
Complement
Linguists commonly borrow terms from algebra to refer to the different parts of phrase structure:
XP
u
Specifier
X’
ru
X
Complement
Just substitute any lexical type for X and you have an X phrase or XP!
An important aspect the so-called X-bar phrase structure is that both the specifier and
complement are phrases in their own right. This feature produces the property of recursion that
allows a rule to be repeated indefinitely. For example, noun phrases have prepositional phrase
complements, and prepositional phrases take noun phrase complements, which creates structures
like:
NP
ru
Spec
N’
ru
N
PP
ru
P
NP
ru
N
PP
|
etc.
e.g., the height of the lettering on the top of the page
Recursion also allows us to add an indefinite number of adjectives to modify a noun:
NP
eo
Spec
N’
qp
AdjP
N’
ty
ep
Spec
Adj
AdjP
N’
ty
t
Spec
Adj
AdjP
ty
Spec
Adj
N
e.g., a very large, very black, rather undignified bulldog
Phrase Structure Grammar
I have been using tree structures to display the syntactic relations between constituents.
Linguists also use phrase structure rules to describe the same relations. A phrase structure rule
looks like the phonological rules we wrote earlier. It contains a symbol on the left side of an
arrow that indicates the upper or mother node in the tree structure, and one or more symbols on
the right side of the arrow that indicate the lower or daughter nodes in the tree structure.
Phrase structure rules and tree diagrams are equivalent notational devices. You can convert one
into the other.
We can start by writing phrase structure rules that generate our first example sentence:
Any speaker of English can split this sentence into two main constituents.
The sentence has two main constituents, a subject and predicate, so we will need a phrase
structure rule like:
S ® NP AuxP
We can use X-bar theory to help us devise a phrase structure rule for the subject. We will need a
specifier and a complement phrase:
NP ® Spec N’
N’ ® N PP
Note that I am assuming the prepositional phrase serves as a complement to the head of the noun
phrase. The other constituent phrases can be divided into heads, specifiers and complements as
follows:
AuxP ® Aux VP
VP ® V NP PP
PP ® P NP
These rules constitute a phrase structure grammar for (a fragment of) English. The rules
describe the constituent structure of our sentence, and also provide a recipe for generating that
sentence and many more. In theory, we should be able to add new rules to our grammar until we
obtained a complete description of the language. Such a grammar would generate every possible
sentence in the language, and would not generate any unacceptable sentences. Despite 50 years of
serious grammar writing, linguists still haven’t produced such a grammar. We do not yet know if
this problem can be solved.
You may have noticed some problems with the phrase structure rules I wrote to generate our first
sentence. As written, the rules do not generate the last noun phrase ‘two main constituents’. How
would we need to change the rule for noun phrases to generate this phrase?
Another problem with the rules is that they do not allow for the option of not generating some
constituent. Three of the noun phrases in our sentence have specifiers, but one does not. One of
the noun phrases has a complement, but three do not. Linguists use parentheses in phrase
structure rules to capture this feature of optionality. We can rewrite the noun phrase rules to
allow for optional specifiers and complements.
NP ® (Spec) N’
N’ ® N (PP)
The option of adding a modifier to the noun phrase raises the need for another type of option if
we add a rule for adjectival modifiers as
N’ ® AdjP N’
Linguists uses curly braces in phrase structure rules to capture this option.
N’ ® N (PP)
9 AdjP N’ A
We face two additional problems that X-bar theory creates. The phrase structure rules to generate
an X-bar phrase would be:
XP ® (Spec) X’
X’ ® X (Comp)
Our new set of rules for noun phrases provide a good example of the X-bar template. Our rule for
the verb phrase is another story. Our verb phrase deviates from the X-bar template in that the
verb has two complements rather than the single complement predicted by X-bar theory. We face
a direct challenge to the theory and have to decide whether we want to keep the theory and find
some way to shoehorn the verb phrase into the X-bar template, or change the X-bar template to
allow multiple complements.
The second problem X-bar theory creates is seen in our first phrase structure rule.
S ® NP AuxP
How does this rule violate X-bar theory, and how can we change the rule to avoid such a
violation?
Lexical Categories
Our phrase structure grammar still lacks one final rule for generating our sample sentence. We
need a way to translate the terminal strings of our rules into actual words. The solution is simple
enough; assume another rule translates the terminal strings into words of the proper lexical
category.
N ® speaker, sentence, constituents
V ® split
Aux ® can
Spec ® any, this, two
P ® of, into
Adj ® main
We can combine all of these rules into a single rule
X ® lexemex
Here, lexeme points to a word in the lexicon that belongs to the lexical category X. This rule
captures the idea that each word belongs to a part of speech that selects the appropriate specifier
and complement phrases.
We can use several types of evidence to detect lexical categories in a language.
Notional categories
We can use word meaning to divide words into various notional categories.
Noun - the name of a person, place or thing
Verb - the name of an action or state
Adjective - the name of a quality
Preposition - the name of a path or location
The problem with such notional categories is that they are rather imprecise. For example is an
action a thing or an action?
Inflectional categories
We can use lexical inflection as evidence for separating words into different lexical categories.
Noun - inflect for plural, possession
Verb - inflect for tense, aspect, voice, mood, agreement
Adjective - inflect for comparative, superlative
The problem with inflectional evidence is that it is not available for every category, and some
words lack an overt inflection.
Syntactic categories
We can also use syntactic frames as evidence for grouping words into lexical categories. Words
are used in specific syntactic contexts, which aid in analyzing the lexical categories.
Noun
Verb
Adjective
Preposition
Auxiliary
Det ______, Det Adj _______
Aux ______, Please _______
______ N, Adverb ________
______ NP, right _______
______ V, ______ not
Other Languages
Our lexical category tests do not apply to other languages. We cannot use inflectional evidence in
languages that lack inflection–analytic languages. Likewise, we cannot use the presence of
determiners as evidence for a noun category if the language does not require determiners in its
noun phrases. We can divide words in any language into notional categories, but then we have no
basis for constructing syntactic rules from notional categories.
We can determine what lexical categories exist in other languages and then try to decide the
extent to which we find the ‘same’ lexical categories in different languages.
One would think that all languages would at least contain the lexical categories of noun and verb.
The Salish languages of the Northwest challenge this assumption.
Kinkade (1983) Lingua, ‘Salish evidence against the universality of ‘noun’ and ‘verb’’
Upper Chehalis only has two lexical classes: predicates and particles
s-q’wct’-w-n
continuative-burn-intrans-3sing
‘fire’ or ‘it is burning’
s-x/ cp-w-n
continuative-dry-intrans-3sing
‘it is drying’
s-º§aláš
‘deer’ or ‘it is a deer’
§it-q’walán’-…
completive-ear-2sing
‘you are all ears’
Columbian Salish
s-q’á§-xn
continuative-wedge_in-foot
‘shoe’ or ‘it is a shoe’
The category of adjectives also varies across languages. Quechua does not distinguish between
adjectives and nouns:
chay runa hatun (kaykan)
that man big
is
‘That man is big’
chay runa alkalde (kaykan)
that man mayor is
‘That man is mayor’
chay hatun runa
‘That big man’
chay alkalde runa
‘That mayor man’ (that man who is mayor)
K’iche’ expresses some adjectival notions as verbs:
x-in-war-ik
x-in-kos-ik
comp-1sing-sleep-status
‘I slept’
comp-1sing-tire-status
‘I was tired’
x-in-noj-ik
comp-1sing-full-status
‘I was full’
x-in-kikot-ik
comp-1sing-happy-status
‘I was happy’
Some languages (Jakaltek) lack prepositions.
Some languages have lexical categories that are not found in English.
Mayan languages distinguish a category of positionals from other lexical types:
k-in-tak’-e§-ik
k-in-paq-e§-ik
incomp-1sing-stand-POSITIONAL-status
‘I am standing’
incomp-1sing-climb-POSITIONAL-status
‘I am climbing’
We have the following categories in languages
Nootka (N, V)
Korean (N, V, P)
Jakaltek (N, V, A)
English (N, V, A, P)
Linear Order
The linear order of heads, specifiers and complement phrases also varies across languages. We
can describe these differences easily by means of phrase structure rules.
Korean (Head Last)
S ® NP VP
VP ® (PP) (NP) V
NP ® (Spec) N
PP ® NP P
Chun ku chayk poata
Chun that book see
‘Chun saw that book’
Selayarese (Head First)
S ® VP NP
VP ® V (NP) (PP)
NP ® N (Spec)
PP ® P NP
Subcategories
la§allei doe§ iñjo ri lamari iñjo i Baso
take money the in cupboard the Baso
‘Baso took the money in the cupboard’
We often need to make finer distinctions between the words in a lexical category. Transitive
verbs allow an object complement, but intransitive verbs do not. Ditransitive verbs take two
complement phrases. There are also verbs that take adjective and sentence complements.
Intransitive verbs (Vi)
Transitive verbs (Vt)
Ditransitive verbs (Vdt)
Adjective complement verbs
Sentence complement verbs
fall, sleep, sneeze, jog
throw, cut, acknowledge, seek
give, show, put
be, become
want, promise, say
Other parts of speech have similar restrictions on the types of complements they take.
Nouns
None
PPto
PPabout
car, electricity
presentation, pledge
argument, discussion
Adjectives
None
PPabout
PPto
tall, smart
curious, glad
apparent, obvious
In addition to complement types there are other types of subcategory distinctions. Singular count
nouns require a determiner, but singular mass nouns do not.
Count nouns table, boy, wind, mesa
Mass nouns water, gold, furniture
Trying to force all words into a specific lexical subcategory creates problems. Large classes of
verbs regularly alternate between specific subcategories. The verbs break, tear and drop as well
as the verbs eat, drink and see can appear as both transitive and intransitive verbs. English
speakers differ over whether the noun data is mass or count. Linguists are currently attempting to
describe and account for a wide range of phenomena associated with lexical subcategories.
Transformations
In addition to describing the internal structure of sentences syntax also looks at the relations
between sentences. For example, we can transform a declarative sentence in English into a
simple question by moving the auxiliary verb to the beginning of the sentence.
Any speaker of English can split this sentence into two main constituents.
becomes
Can any speaker of English split this sentence into two main constituents?
If we maintain that these two sentences are related in some fashion, then we can try to account
for the relation. We could propose a movement rule, or transformation, to account for the
relation. Such a rule raises a number of issues.
First Guess
Inversion Transformation
X Aux Y ® Aux X Y
One problem with this version of the rule is that it is overly general. Think of an unacceptable
sentence that this rule generates.
Another problem with this rule is that it doesn’t explain why we move the auxiliary to the
beginning of the sentence. A subpart to this problem is that the rule does not tell where to put the
auxiliary once we move it.
We can solve both problems by assuming the Auxiliary moves to a position that distinguishes the
illocutionary force of a question from that of a declarative sentence. We can call the node that
heads this position the Question node, which projects a Question phrase. I am assuming the
process of forming a question includes creating a Question phrase which then attracts the
auxiliary verb. We can make a further assumption that the auxiliary verb lands at the head of the
Question phrase. Our revised inversion rule would then be:
Second Guess
Inversion Transformation
Question X Aux Y ® Question-Aux X Y
The variable X in this formulation is still too general. We can tighten the rule further by
specifying that X always refers to a subject.
Third Guess
Inversion Transformation
Question NP Aux Y ® Question-Aux NP Y
This version of the rule asserts that the first auxiliary after the first noun phrase at the beginning
of the sentence is the word that moves to the Question Phrase.
Other Languages
Looking at other languages we find some interesting differences in their inversion rules.
French allows an auxiliary or main verb to move to the Question node.
Vois tu ___ le livre?
see you ___ the book?
‘Do you see the book?’
As tu ___ essayé?
have you ___ tried?
‘Have you tried?’
French only allows a verb or auxiliary to move across pronominal subjects. Movement across
non-pronominal subjects is not acceptable.
*Sait Jean?
‘Knows John?’
*A Jean essayé?
‘Has John tried?’
French speakers resort to one of the following constructions to form such questions.
Est-ce que Jean sait?
‘Is it that John knows?’
Jean sait-il?
‘John knows-he?’
Spanish moves the main verb to the Question node, but not auxiliary verbs. The rule applies
regardless of whether or not the subject is a pronoun.
Partió Juan ___ ?
leave John ___ ?
‘Did John leave?’
Partió él?
‘Left he?’
*Ha Juan ___ partido?
has John ___ left?
‘Has John left?’
*Ha él partido?
‘Has he left?’
K’iche’ simply introduces a particle to mark questions.
A xel lee a Waan?
Q left the familiar John?
‘Did John leave?’
The K’iche’ rule resembles what happens in English when the underlying sentence does not
contain an auxiliary verb. English introduces the auxiliary do is such situations, giving rise to a
rule known as Do-Support. The French est-ce que construction suggests the language is moving
in the direction of K’iche’.
The English Inversion rule changed during the sixteenth and seventeenth centuries. Before this
time English speakers could invert the main verb.
Speak they the truth?
Cross-linguistic comparison reveals several parameters are necessary to account for inversion
across languages. All of these languages mark yes-no questions by marking a question feature at
the left periphery of the sentence. The parametric variation for inversion includes:
1. Whether or not a particle is added to mark the question.
2. Whether or not the main verb moves.
3. Whether or not the auxiliary moves.
4. Whether or not the verb can move across non-pronominal subjects.